Let the Denial BeginPosted: October 7, 2012
It is an awful fact that women are very underrepresented in my discipline, Computer Science, and as an aggregate across my faculty, which includes Engineering and Mathematics (so we’re the Technology, Engineering and Mathematics of STEM). I have heard almost every tired and discredited excuse for why this is the case but what has always angered me is the sheer weight of resistance to any research that (a) clearly demonstrates that bias exists to explain why this occurs, (b) identifies how performance can be manipulated through preconceptions and (c) requires people to consider that we are all more similar than current representation would indicate.
Yes, if I were to look around and say “Women are not going to graduate in large numbers because I see so few of them” then I would be accurate and yet, at the same time, completely missing the point. If I were to turn that around and ask “Why are so few women coming in to my degree?” then I have a useful question and, from various branches of research, the more rocks we turn over, the more we seem to find bias (conscious or otherwise) in both industry and academia that discourages women from participation in STEM.
A paper was recently published in the Proceedings of the National Academy of Sciences of the United States of America (PNAS, to its friends), entitled “Science faculty’s subtle gender biases favor male students”. (PNAS has an open access option but the key graphs and content are also covered in a Scientific American blog article.) The study was simple. Take a job application for a lab manager position. Assign a name where half of the names are a recognisably male name, the other half are female. (The names John and Jennifer were chosen for this purpose as they had been pre-tested to be equivalent in terms of likability and recogniseability.) Get people to rate the application, including aspects like degree of mentoring offered and salary.
Let me summarise that: the name John or Jennifer is assigned to the same application materials. What we would expect, if there is no bias, is that we would see a similar ranking and equivalent salary offering. (All figures from the original paper, via the SciAm link.)
Oh. It appears that the mere presence of a woman’s name somehow altered reality so that an objective assessment of ability was warped through some sort of … I give up. Humour has escaped me. The name change has resulted in a systematic and significant downgrading of perceived ability. Let me get the next graph out of the way which is the salary offer.
And, equally mysteriously, having the name John is worth over $3,500 more than having the name Jennifer.
I should leap to note that it was both male and female scientists making this classification – which starts to lead us away from outright misogyny and towards ingrained and subtler prejudices. Did people resort to explicitly sexist reasoning to downgrade the candidates? No, they used sound reasoning to argue against the applicant’s competency. Except, of course, we draw back the curtain and suddenly reveal that our sound reasoning works one way when the applicant is a man, another if they are a woman.
Before you think “Oh, they must have targeted a given field, age group or gone after people who do or don’t have tenure”, the field, age and tenure status of the rating professors had no significant effect. This bias is pervasive among faculty, field, age, gender and status. The report also looked at mentoring and, regardless of the rater’s gender, they offered less mentoring to women.
Let’s be blunt. Study after study shows that if there are any gender differences at all, they are so small as to not even vaguely explain what we see in the representation of female students in certain fields and completely fails to explain their reduced progress in later life. However, the bias and stereotypes that people are operating under do not so much predict what will happen as shape what will happen. We are now aware of effects such as Stereotype Threat (Wiki link) that allows us to structure important situations in someone’s life so that the framing of the activity leaves them in a position where they reinforce the negative stereotype because of higher anxiety, relative to a non-stereotyped group. As an example, look at Osborne, Linking Stereotype Threat and Anxiety, where you can actually reduce the performance of girls on a maths test through reminding them that they are girls and that girls tend to do worse on test than boys. Osborne then compared this with a group where the difference was identified but a far more positive statement was made (the participants were told that despite the difference, there were situations where girls performed as well or better). The first scenario (girls do worse) was a high Stereotype Threat scenario (high ST), the second is low ST. Here’s the graph from Wikipedia that is a redrawing of the one in the paper that shows the results.
That is the impact of an explicit stereotype in action – suddenly, when framed fairly and without an explicit stereotype or implicit bias, we see that people are far more similar than we thought. If anything, we have partially inverted the stereotype.
To return to my first paragraph, I said:
what has always angered me is the sheer weight of resistance to any research that (a) clearly demonstrates that bias exists to explain why this occurs, (b) identifies how performance can be manipulated through preconceptions and (c) requires people to consider that we are all more similar than current representation would indicate.
The PNAS paper, among others, clearly shows that the biasses exist. A simple name change is enough, as long as it’s a woman’s name. The demonstrated existence of stereotype threat shows us how performance can be manipulated through preconception. (And it’s important to note that stereotype threat is as powerful against minorities as woman – anyone who is part of a stereotype can be manipulated through their own increased or reduced anxiety.) So let me finally discuss the consideration of all of this and the title of this post.
I am expecting to get at least one person howling me down. Someone who will tear apart all of this because this cannot, possibly, under any circumstances be true. Someone who will start talking about our “African ancestors” to start arguing the Savanna-distribution of roles, as if our hominid predecessors ever had to apply to be a lab manager anywhere. Most of you, I hope, will read this and know all of this far too well. Some of you will reflect on this and, like me, examine yourself very carefully to find out if you have been using this bias or if you have been framing things, while trying to help, in a way that really didn’t help at all.
Some of you, who are my students, will read this and will see that research that you have done is reflected in these figures. Yes, we treat women differently and we appear, in these circumstances, to treat them less well. This does not, under any circumstances, mean that we have to accept this or, in any way, respect this as an established tradition or a desirable status quo. But the detection of an insidious and pervasive bias, that spans a community, shows us how hard my point (c) actually is.
We must first accept that there is a problem. There is a problem. Denying it will achieve nothing. Arguing minutiae will achieve nothing. We have to change the way that we react and be honest with ourselves that, sometimes, our treasured objectivity is actually nothing of the kind.